ptls.trx_encoder

All classes from ptls.nn.trx_encoder also available in ptls.nn

ptls.nn.trx_encoder helps to make a representation for single transactions.

ptls.nn.TrxEncoder

Now we have an input data:

x = PaddedBatch(
    payload={
        'mcc_code': torch.randint(1, 10, (3, 8)),
        'currency': torch.randint(1, 4, (3, 8)),
        'amount': torch.randn(3, 8) * 4 + 5,
    },
    length=torch.Tensor([2, 8, 5]).long()
)

And se can define a TrxEncoder

model = TrxEncoder(
    embeddings={
        'mcc_code': {'in': 10, 'out': 6},
        'currency': {'in': 4, 'out': 2},
    },
    numeric_values={'amount': 'identity'},
)

We should provide feature description to TrxEncoder. Dictionary size and embedding size for categorical features. Scaler name for numerical features. identity means no rescaling.

TrxEncoder concatenate all feature embeddings, sow output embedding size will be 6 + 2 + 1. You may get output size from TrxEncoder with property:

>>> model.output_size
Out[]: 6

Let's transform our features to embeddings

z = model(x)

z is also PaddedBatch. z.seq_lens equals x.seq_lens. z.payload isn't dict, it's tensor of shape (B, T, H). In our example B, T = 3, 8 is input feature shape, H = 6 is output size of model.

Now we can use other layers which consume transactional embeddings.

Classes

See docstrings for classes:

• ptls.nn.TrxEncoder